Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Deb Schrag, MD, MPHChief, Population Sciences Senior Physician, Gastrointestinal Cancer CenterDana-Farber Cancer InstituteProfessor of Medicine, Harvard Medical School
• Personal fees from JAMA for editorship
• Research support from:• GRAIL• Pfizer
• Collaboration on RWD with:• AACR GENIE• FDA• Cancer-LINQ• Flatiron• NCI-SEER
• Reasons for intense interest in RWD/RWE
• Use cases for RWD/RWE
• Challenges in using RWD/RWE
• Expanding capacity for RWD/RWE use
• Act passed by US Congress in 2016• Intensive lobbying by patients and industry• Focuses on improving access to medicines and accelerating research• The Cures Act requires FDA to establish a program to evaluate RWE to:
• Support the approval of a new indications for a drug approved under section 505(c) of the Federal Food, Drug, and Cosmetic Act
• Support or satisfy post-approval study requirements
• FDA released draft framework with guidance for industry in December 2018
• RWD is defined as: data relating to any aspect of patient health status that are collected in the context of routine health care delivery
• RWD come from a variety of sources including:• Billing claims and encounter records• Population-level registries • Disease and device registries• Electronic health records• Patient-reported data• Electronic surveillance data (activity trackers, implants, wearables)• Large pragmatic trials—eg cluster RCTs
• RWD is defined by what it is not----data from prospective RCTs
• RWD’s purpose is to generate Real World Evidence
Drug Development is lengthy and Failure Rates are High
Public demand for accelerated progress
Problems with the traditional paradigm
Too few participants means that RCT results:
• Have high internal validity
• Often lack external validity
• Do not generalize to typical patients
Genomic Data
What features?
Outcomes Data
Achieve best results?
Therapeutic Data
Which treatments?
• <5% of patients are treated on a clinical trial• To learn from the other 95%, we need to focus on
strengthening the RWD ecosystem
Venook et al JAMA 2017
• Take a long time
• Require many patients
• Fail to detect small subsets that derive benefit
• Growth and dissemination of Electronic Health Records
• International data standards that facilitate meaningful data sharing
• Growth in computational power• Fast computers• Declining costs for data storage
• Reasons for intense interest in RWD/RWE
• Use cases for RWD/RWE
• Challenges in using RWD/RWE
• Expanding capacity for RWD/RWE use
• To support drug development, testing and regulatory approval
• Relatively new• Controversial
• After a drug receives regulatory approval from FDA/EMA or other regulatory authority
• Quite standard • Use cases are expanding
• RWD in the Post Drug Approval Space
• Meet post-marketing commitments
• Pharmacovigilance
• Identify exceptional subgroups
• Monitor dissemination
• Measure efficacy-effectiveness gap and public health impact
• Expand drug label to new indications
TRADITIONAL APPROACH
• Commitments often not met• Expensive for industry• Often have low scientific yield• Often in select populations• Capture pre-specified outcomes on
case report forms
RWD APPROACH
• Use standardized EHRs to:• Evaluate treatment experience for all
drug recipients• Capture outcomes that can be
extracted from the EHR• Dosing, dose modifications• Major or lab toxicities• Health system use-ER/Hospitalizations• Measure time to treatment discontinuation• Measure overall survival• Challenging to measure response, DFS
TRADITIONAL APPROACH
• Rely on voluntary reports from users—difficult to sift through and inconsistently submitted
• Rely on firms to meet post-marketing commitments in timely fashion
• Minimal patient reported data
RWD APPROACH
• Use standardized EHRs to
• Interrogate lab data for all users• Measure standardized outcomes including
hospitalizations, dose delays, dose reductions
• Use Patient Reported Outcomes• Measure toxicity from patient perspective• Understand the patient experience
TRADITIONAL APPROACH
• Extensive analysis of RCT patient data subsequent to drug approval to identify “exceptional responders” or “exceptional resistance” as well as severe and/or mild toxicity
• Focus on select subgroups in Phase IV trials at select study sites
RWD APPROACH
• Use standardized EHRs combined with routinely performed molecular profiles to identify very small molecular subsets
• Measure outcomes in all patients with available EHR data
• Creative and strategic data partnerships• Foundation Medicine-genomic data• Flatiron Health-EHR data curation
TRADITIONAL APPROACH
• New drug labeling indication requires new clinical trial demonstrating benefit
• Many trials in niche populations never performed
• Clinicians left to extrapolate and make decisions to the extent permissible by regulatory authorities
• Decisions get made but knowledge not captured for use in future cases
• Appendix cancer?• Duodenal cancer?• Her2+ colorectal cancer?
RWD APPROACH
• Capture “off label” use and measure outcomes in consistent fashion from EHR data
• Use data to expand/contract the label
• Label then used to dictate coverage and/or reimbursement levels
• FDA expanded palbociclib label to include men with HER2+/ER- breast cancer in 2019
• ~2700 cases of male breast cancer per year in US• RCTs would be challenging• PALOMA2 and PALOMA3 excluded men
• Multiple data sources• Data from company registry (Pfizer)• EHR data from Flatiron• Insurance claims data from IQVIA
• RWD in the Pre-Approval Space• Identifying clinical need• Selecting optimal sites to conduct study• Informing study design• Inclusion/exclusion criteria• Realistic timeline for accrual• Realistic target endpoints and appropriate sample size
• Clinical trial execution• Simplifying case report forms• Decreasing burden and expense of data collection• Facilitate interoperability across studies
• Support applications for accelerated approval based on uncontrolled studies• Avoid the need for randomization by using “synthetic control” arm from RWD
• Blinatumomab received approval for Philadelphia chromosome-negative relapsed and refractory B-cell precursor acute lymphoblastic leukemia based on a single phase II trial
• The results were compared with historical data from 694 comparable patients extracted from 2,000 patient records in the U.S. and E.U
• Avelumab was approved by the FDA in 2017 for treatment of metastatic Merkel Cell Carcinoma (~1600 cases per year in the USA)
• Phase II data demonstrated significant and durable response
• RWD describing outcomes in historical cohorts of patients were included in the application
• Ideal use case for RWD:• Very rare cancer• No previous effective treatment• Very clear strong signal from phase II• Very homogeneous treatment in RWD• Clear reproducible endpoint (Overall Survival)
• Extremely challenging to use “synthetic controls” for common cancers• Validity traps abound• Unlikely to be acceptable to regulators absent very strong signal• Observational data cannot control for confounding by unmeasured factors• Trials with endpoints other than survival are especially challenging• RCTs will remain the backbone of progress• Strategies to improve execution and conduct of RCTs must be prioritized
• Reasons for intense interest in RWD/RWE
• Use cases for RWD/RWE
• Challenges in using RWD/RWE
• Integrating Clinical Trials and RWD/RWE
• Data privacy and security concerns
• Many obstacles to data sharing
• Poor interoperability across and within systems
• Lack of international standards for measuring the most salient outcomes• RECIST is a standard for measuring PFS in RCTs• RECIST isn’t workable in real world contexts• No standards for measuring DFS, PFS, response from unstructured EHR data
• Observational data from study of MSI vs MSS patients treated with or without chemotherapy
• Appears that MSI patients obtain greater benefit from chemotherapy
• Misleading because MSI patients are younger and healthier
• Very difficult to remove selection bias
Elsaleh Lancet 2000
THEN: Paper Records NOW: 100% EHRs
Same old problem:Uninterpretable by computers;Unusable for research
Unstructured Data
Immense information to be gleaned from unstructured data but it requires intelligence to analyze and interpret
Structured DataUnstructured Data
Even in the best EHRs, only a small % of “big data” is structured
WE NEED TO UNLOCK THE BLACK BOX
Today, EHRs are a black box Most of the information they contain is not structured and therefore not interpretable
PRISSMM is a standard taxonomy for classification and communication of structured information about cancer status and treatment outcomes following the assignment of TNM stage for patients with solid tumors
Each letter in PRISSMM corresponds to a dimension of cancer status or treatment response.
PRISSMM: A Taxonomy for Defining Cancer Outcomes
Pathologic evidence of locoregional or distant evidence of tumor
Radiographic evidence of locoregional recurrent or persistent tumor
Imaging evidence of distant/disseminated tumor beyond the primary site
Signs of cancer on physical exam or symptoms that can be attributed to tumor
Symptoms of tumor on physical exam or symptoms that can be attributed to tumor
Tumor Marker evidence of persistent or recurrent tumor
Oncology Medical Provider assessment
Each curation effort may focus on some or all of the PRISSMM componentsSigns may be relevant for melanoma outcomesMarkers may not be relevant for lung outcomes
D. Schrag
• Expedites structuring unstructured data (gold standard)• Enables interoperability and data sharing across sites• Streamline EHR abstraction by research staff/registrars• Facilitates “computer assisted” abstraction• Breaks abstraction/annotation into discrete tasks for machines• Creates a “lingua franca” that is understood by the clinical and
research community• Allows customized definitions of PFS/DFS from “real world” data• Data standard works for all solid tumors
What Problems Does PRISSMM Aim to Solve?
PRISSMM: Curation of Individual Patient TrajectoryColon Cancer Stage IV, First Treatment [℞1]
PRISSMM: Curation of Individual Patient TrajectoryReal World Progression-Free Survival [Anchored from Diagnosis: Dx]
PFSI: Dx-13
PFSMD: Dx-16
PFSI or MD: Dx-13
PRISSMM: Curating Patient OutcomesReal World Progression-Free Survival: Anchored from Treatment [℞]
PFSI: ℞1-12
PFSMD: ℞1-15
PFSI or MD: ℞1-12
PRISSMM CurationStage IV Colon Cancer
• Reasons for intense interest in RWD/RWE
• Use cases for RWD/RWE
• Challenges in using RWD/RWE
• Integrating Clinical Trials and RWD/RWE
West H, JAMA Oncology 2017
Novel designs mandate strategic use of RWD
Impossible to execute these trials without a robust data ecosystem that efficiently permits access to RWD
Just because vemurafenib has efficacy for BRAF+ melanoma doesn’t mean it will have efficacy for
BRAF+ colorectal cancer
Drillion et al NEJM 2018
Genomic Data Phenomic DataTherapeutic
Data
Knowledge Integration to Accelerate Discovery
Important Strategies:• Engaging patients directly: PROs• Artificial intelligence and Machine Learning
ProteomicsEpigenetics
Germline DNAPatient Reported
Data
• RCTs will:
• Become more varied in design
• Innovative designs depend on nimble data ecosystem
• Will rely on RWD to become more strategic across the life cycle
• RWD will:
• Grow and expand rapidly
• Depend on ability to overcome non-technical barriers
• Never completely replace RCTs